Public Opinion Data over Time and across Countries

This chapter draws on public opinion data spanning more than sixty years in the United States, the United Kingdom and Canada, over fifty years in Germany and some twenty-five years in Australia. This comprises a total of 10,920 individual survey items relating to issue competence. These aggregate data are collected from a range of sources.

Our US data were compiled from the database of survey data held by the Roper Center for Public Opinion Research. Our UK data are drawn from a combination of Gallup polls, as reported in King and Wybrow’s (Reference King and Wybrow2001) British Political Opinion 1937–2000; the online archives of the polling firms Ipsos MORI, YouGov and Populus; a dataset compiled as part of the Continuous Monitoring Survey (CMS) between 2004 and 2013; and surveys of the British Election Study between 1963 and 2010. The Australian survey data are from the online archives of Newspoll. The Canadian data were compiled from Bélanger (Reference Bélanger2003) and original monthly reports of the Canadian Gallup Index held at McGill University. Our data for Germany come from a combination of the GESIS archive holdings of German Election Studies and Forschungsgruppe Wahlen’s monthly Politbarometer from 1977 to 2013, supplemented with additional data from the online archives of Forschungsgruppe Wahlen.

Survey questions aimed at measuring public evaluations on party handling of issues take a range of forms. Common formulations ask respondents ‘which party’ would handle issue X ‘best’, or ‘would do a better job’ handling issue Y, or which would ‘you trust to do a better job on’.

I am going to read out a list of problems facing the country. Could you tell me for each of them which political party you personally think would handle the problem best? Education and schools

With Britain in economic difficulties, which party do you think could handle the problem best – The Conservative Party or the Labour Party?

Do you think the Republican Party or the Democratic Party would do a better job of dealing with each of the following issues and problems?

Which political party, the Democrats or the Republicans, do you trust to do a better job handling … the economy?

Which one of the (ALP, Liberal and National Party Coalition or someone else) do you think would best handle welfare and social issues?

There are cross-national and within-country variations in question wording but all relate to some aspect of handling, competence, performance, effectiveness, trust or delivery. These tend to tap citizens’ evaluations of party competence in similar ways. A typical question format in the United States asks respondents ‘Who do you trust to do a better job of handling the economy: the Democrats or the Republicans?’ or ‘Do you think the Democratic Party or the Republican Party can do a better job in … reducing the crime rate … or don’t you think there’s any difference between them?’ The most common question format in the United Kingdom asks respondents ‘I am going to read out a list of problems facing the country. Could you tell me for each of them which political party you personally think would handle the problem best?’ In Germany, a variant of these asks, ‘Welche Partei ist Ihrer Meinung nach am besten geeignet, neue Arbeitsplätze zu schaffen?’ [Which party is best, in your opinion, at creating new jobs?].Footnote ¹ In Canada, the question is posed along the lines of ‘Which federal political party do you think can best handle the problem of social security?’ In Australia, pollsters ask ‘Which one of the (ALP, Liberal and National Party Coalition or someone else) do you think would best handle national security?’ The survey items, therefore, typically relate to which party is best able to handle or deliver on a particular issue, relative to other parties.

Due to the extended time span of the data, these consist of a variety of survey modes (face-to-face, telephone, online panels). Most refer to all voting-age adults. While the US data does include some items with registered or likely voter filters, this is not the norm. The sample sizes can also vary, but they are typically in the region of 1,000 or more, with just a few instances where polling firms use split samples to ask about competence across a greater number of issues (such as for the NBC/Wall Street Journal poll series regarding party issue handling).

In an ideal world we would use survey data collected using a consistent weighting and sampling strategy, but in the absence of this luxury the data provide the best available measures of how each party is perceived as capable of handling a particular issue at any given point in time. To the extent that weighting and sampling impacts on these measures it serves to introduce error – and thus, if anything, our analyses will understate the importance of issue competence.

One option for analysing these data would be to examine issue evaluations recorded by a single source (e.g. the same polling house over an extended period of time) where ratings of parties across issues should be expected to be most similar (respondents often rate parties across a range of issues in the same survey and so the influence of partisanship and of satisficing in survey answers will be maximised). Another is to evaluate survey data gathered across a range of different sources, examining greater amounts of data on more issues and over a longer time period with fewer gaps. We opt for the latter approach, taking the average ratings of parties on issues over multiple survey items in a given year. We aggregate our data within particular issue topics so that trends can be compared over time and across countries. Survey questions on issue handling are often fielded irregularly and infrequently by polling organisations and national election studies. We develop a classification scheme, adapted from the Comparative Agendas Project, consisting of thirteen issues. These are issues that attract the attention of mass publics, media, parties and governments; the major issues on the public agenda. Most correspond directly to specific domains of public policy, such as the economy, health, defence and education. The classification scheme is outlined in Table 3.1, with corresponding examples for each issue category. Managing public spending, reducing taxes and balancing the budget are classified as economic issues. Dealing with pensions and social security are welfare issues. National security and terrorism are defence issues. The exception is our thirteenth topic, morality, which relates to moral standards and promotion of family issues, another focus of parties and publics (see Engeli et al. Reference Engeli, Green-Pedersen and Thorup2012).

This approach to classification enables us to consider how evaluations of issue-handling change or remain stable over time. It also allows for comparative insights. It does mean that evaluations of parties on these aggregate issues might fluctuate due to the evolution of the issue topic – for example, as the focus of defence changes from being about the Cold War to preoccupation with terrorism. Likewise, the historical focus of rights and minorities on civil rights and race relations, in the United States, has more recently incorporated rights as they relate to same-sex marriage and gender. To study ownership, performance and generalised competence over time we have to allow that the underlying mix of issues can adapt and change. By aggregating across survey items, we expect substantial changes in issue competence to therefore represent meaningful change; not simply change that cancels out within issue categories.

Public Opinion about Party Competence

We begin our analysis by displaying data on public evaluations of the major parties from our dataset.

In Figure 3.1, we display aggregate-level data on the public’s evaluations of the US Democratic Party on the thirteen issues for which we have data over time, and in Figure 3.2 for selected parties in the United States, the United Kingdom, Australia and Germany for a smaller subset of issues. This enables closer visual inspection of trends in evaluations of party handling of individual issues.Footnote ² There are not enough survey data for Canada to enable inspection of party issue competence for individual issues, and for Germany we have only four issues with sufficient data. The figures cover the periods for which the data are richest (and thus most informative) between 1980 and 2013 in the United States, between 1980 and 2015 in the United Kingdom, between 1989 and 2014 in Australia and between 1980 and 2013 in Germany.

Figure 3.1 Public opinion about issue handling for the US Democratic Party, 1980–2013

Figure 3.2 Public opinion about issue handling for parties in the United States (a), the United Kingdom (b), Australia (c) and Germany (d)

The trends in Figure 3.1 exhibit three immediately clear features. The first is that the Democrats have no issues that are clearly and stably owned over time. Their best issue in the earlier time period is welfare and housing, but this is later replaced by the environment (in around 2000). The second feature is the amount of noise and variation in ratings on individual issues. Note that by aggregating sub-topics into issue categories we are already dampening some of the noise and variance by design. Still there is substantial variance we need to explain. This is not all subsumed in economic ratings (the bold black line), which remain almost constantly in the middle of owned issues (best rated) and not owned issues (worse rated). The third feature is the shared over time variation we observe for the Democrats. There are clear peaks and troughs for all issues as ratings move upwards and downwards in common, while retaining distinct level differences, a shared improvement in Democrat handling evaluations and a narrowing in the distance between Democrat ratings across issues from about 2005.

To enable better inspection of individual issue patterns, we can now zoom in on a selection of issues across other cases.

These figures display a number of features in common, as well as some interesting differences.

There appears to be less issue-specific fluctuation in the ordering of issue evaluations in the United Kingdom in comparison to the United States, especially in the earlier part of the series, and much less so in Australia. The Australian data are drawn from a single polling source and over a shorter time period, so this may explain this difference, but it may also arise due to the highly stable party system in Australia.Footnote ³ The United States and United Kingdom differences may also arise due to the different volume of data across both countries and the degree to which the data are drawn from multiple or single sources. However, the possibility of country and institutional differences, as well as differences over the time period in both countries, is noticeable in Figure 3.2.

There are some hints in Figures 3.1 and 3.2 of increasing homogeneity of issue-handling evaluations over time in both the United States and United Kingdom. We see greater variation for the Democrats in Figure 3.1 in the earlier years, and some narrowing for the Republicans and less variance for the Conservatives in the latter periods in Figure 3.2. Given that we only focus in Figure 3.2 on a selection of issues and parties, we may be missing part of the story. But the full figures also back up this apparent change (see the Appendix, Figures A3.1 to A3.8). We have more data in our later time periods, so this may provide part of the explanation. It is also possible that parties in these countries are losing clear associations with issues.

There is a more complex picture in the data revealed in Figures 3.1 and 3.2 than explained by the traditional notion of issue ownership as a stable feature of party reputations. Each party has a set of issues that tend to be better rated for a period, and some issues that tend to be worse. However, there is also over-time variability on the top issues and bottom issues. In addition to the Democrats’ gain of the environment (and apparent loss of welfare and housing), the Republican ownership of defence in Figure 3.2 appears to substantially weaken by 2010. The same loss can be seen for the UK Conservative Party, but on the issue of foreign affairs. Even the issue of healthcare for the Conservatives becomes less of a clear liability for the party at this point than the issue was between 1980 and 2000. Whereas we see stability of ownership in Australia in Figure 3.2, even only using four issues in Germany we find evidence of ownership change. The German right-of-centre Christian Democratic Union (CDU) seems to gain the issue topic of labour and employment relative to their earlier ratings on this issue. Note that by presenting all issues by party, not between party, we capture variance that is overlooked by examining the lead of one party over another.

In addition to a notion of issue ownership, Figures 3.1 and 3.2 reveal a considerable amount of over-time variation in the average levels at which the public rates each party on issues. Much of this will be random noise, but much will not. The data also reveal that these changes sometimes alter the overall ranking of a party across issues. This point is clearly demonstrated in the Republican ratings on defence in Figure 3.2. Here we see the Republicans retain a relative lead (over other issues) on this issue throughout the period until around 2005 but the party secures significant gains in its ratings in different time points, and incurs significant losses too. This issue-specific variation should be highly important as a party considers how it is evaluated entering an election, or as voters who consider that issue important decide whether to stick with the party or defect.

In addition to relatively stable differences in level, and in addition to issue-specific differences in change, the data in Figures 3.1 and 3.2 reveal that there is a degree to which the public updates its ratings of parties on issues in common. As discussed, the US Democratic Party is evaluated more positively for a time between 2000 and 2010; the party is evaluated more positively on all issues: on higher ranked and lower ranked issues alike. This common shift is not a simple partisan story, as is clear in Figure 3.2. If we ignore the Republicans’ best issue (defence) for much of the series, and its worst issue (environment), there is some indication that the Republicans’ ratings on many other issues have moved in a more positive direction over the entire period, whereby the ratings on issues such as ‘labour and employment’, education, ‘welfare and housing’, morality and ‘rights and minorities’ all climb in common. The common variation is apparent for the UK Conservative Party too. There is a clear structural break in the data which coincides with 1993; this was the year following the events of the infamous ‘Black Wednesday’, in which the Conservatives lost their reputation for economic competence over the ERM crisis. That period also coincided with party disunity, which further affected the party’s reputation for governing, under then Prime Minister John Major. Figure 3.2 makes clear that this reputational damage translated to ratings across policy issues. The party begins to improve its ratings on all issues in common between 2000 and 2010, but the data also reveal some shared peaks in competence: in 1997 and in 2000. The Australian Liberal Party experienced a downward shift in common across issues in 2008 to 2009, a time at which the party was divided over policy (specifically over its response to the Rudd government’s proposed emissions trading scheme) and changed leader twice in a period of just over a year.

Each of these graphs reveals the three concepts we argued for in Chapters 1 and 2. They show how each party has issues on which it tends to be better rated over time and how these within-party better or worse issues can change. Taking the mean of any of these issues, or the lead of one party over another, as measures of ‘ownership’ would confuse two further characteristics: (i) the degree to which every issue exhibits specific variation in some time points or in all and (ii) the degree to which there is a shift in the average ratings across issues for a party, which runs alongside ownership of issues and specific issue volatility. It would also force issue ratings and changes to be, in essence, symmetric, when our data reveal that this assumption is not accurate.

The remainder of this chapter focuses in greater detail on the properties of each of three concepts in turn.

The Characteristics of Issue Ownership

Substantively, issue ownership theory accounts for (a) those issues on which a party should prefer to focus election campaigns and election choices upon, and (b) the relationship between voting behaviour and the salience of issues on which a party is considered the ‘issue owner’. This can explain the focus on the mean level of, or the lead in, positive competence ratings as the empirical focus of attention, since votes will be associated with positive/higher ratings – whether those arise from long-term reputation and representation or whether they arise from short-term fluctuations. However, this operationalisation cannot help us with different theoretical and empirical questions. We seek to explain how issue ownership arises, and how reputations on issues are lost. For this, it is much more meaningful to examine ownership as the long-term reputation of a party for strength and competence on an issue relative to other issues, rather than relative to another party’s rating on an issue.

We propose that ownership is operationalised and measured according to the rank ordering of a party’s issues. This allows us to determine which issues a party should view as its assets, since assets will always be relative to overall performance. It allows one party’s reputation to be damaged on an issue without another party’s reputation being automatically enhanced. It provides insight into important variation in the loss, gain and stability of a party’s ‘best’, ‘worst’ or neutral issues. And, as we show below, it provides a measure that is less susceptible to the problem of endogeneity in ratings of preferred and rival parties on issues.

Taking our dataset of aggregate survey items on issue competence, we create a variable where an issue is the top ranked issue for a party (i.e. it receives a higher score, on average, than every other issue), second ranked issue, third ranked issue, fourth ranked issue and so on, through to the bottom ranked issue (from the n of issues for which survey items are available). We calculate the average score and a ranking by election cycle to ensure that our rankings are not based on too large a number of missing cells (a possibility evident from the incomplete coverage of issues for earlier time periods, before the 1970s, in the United States and the United Kingdom), and also to reflect a long-term notion of ownership which should not change due to measurement error and noise. Such a measure of issue ownership could be calculated for any defined unit of time (such as by year).Footnote ⁴

Tables 3.2 to 3.9 present the within-party ranking of each party’s issue competence over the longest time period for which data are available for the United States, the United Kingdom, Australia and Germany. Additionally, we highlight issues that are in the top-, middle- and bottom-third of a party’s issues depending on how many issues our survey data cover overall (with middle-ranked issues marked in grey text). This summarises each party’s best and worst issues. Having identified each party’s relative issue strengths and weaknesses, we then classify cases of gains, losses or realignments (switching) of ownership, and indicate those in Tables 3.2 to 3.9. Our criteria are as follows.

Table 3.2 presents our cases of ownership and ownership change for the US Republican Party. Table 3.3 follows for the US Democratic Party, Table 3.4 for the UK Conservative Party, Table 3.5 for the UK Labour Party, Table 3.6 for the Australian Liberal Party, Table 3.6 for the Australian Labor Party, Table 3.8 for the German Christian Democrats and Table 3.9 for the German Social Democrats.

(1) Gains are where an issue moves at least two places from the middle-third of a party’s issues into the top-third and it remains there for at least two election cycles (where data are available). This criterion requires that there has to be more than a minor change in ranking and that this change has to persist for a meaningful period of time. This means we do not count cases where a party briefly improves its rankings – perhaps due to a short-term change in evaluations or due to noise in the underlying survey data.

(2) Losses are where an issue moves at least two places from the top-third of a party’s issues into the middle- or bottom-third, remaining there for at least two election cycles. Again this requires the change in ranking to be consequential in its order and persistence.

(3) Lastly, issue ownership realignments are where an issue moves from the top-third of a party’s issues into the bottom-third, or moves from the bottom-third of a party’s issues into the top-third, remaining there for at least two election cycles, and at the same time the issue moves to the top-third or the bottom-third of its opponent’s issues.

Issue ownership gains and losses are indicated by the issues in bold font, with a number reported denoting the number of rankings the issue has moved either positively (e.g. +3) or negatively (e.g. −3).Footnote ⁶

The first thing to note from Tables 3.2 to 3.9 is that there is a considerable amount of variance in issue ownership, even applying our criteria that a change should be a change of at least two rankings and move from a top issue to a middle or bottom issue (or the reverse for a gain), and remain so for two election cycles. Note also that these changes are exhibited in data that have been aggregated across sub-topics and which take an average rating at each period. We are making it difficult to detect a change, and yet we still find considerable variation in party reputations on issues. This is true for the findings in the United States, the United Kingdom and Germany, though not for Australia. This variance is not just seen in issue ownership losses and gains but also in the extent of reputational change. In many instances, an issue moves from being one of a party’s best issues to its worst. An example of this is the issue category of ‘government’ (this includes survey items that refer to running the government efficiently, reforming government and delivering on public services) for the Republicans. It moves from being the party’s top-ranked issue in 1976–1980 to its eleventh-ranked issue in 2008–2012. Another example is the issue of foreign affairs for the Democratic Party, which was the party’s fourth-ranked issue in 1976–1980 but its thirteenth best issue in 2008–2012. The Conservative Party’s rating on ‘labour and employment’ moves from second place (between 1987 and 1997) to tenth place in 2001–2005.

The second thing to note is that few of the changes in ownership are instances where one party’s loss of an issue results in another party’s gain. The Republicans lose the issue of ‘government’, for example, by 1992. This issue only enters the top four issues for the Democrats during the 2012 to 2016 election cycle (and note this data point is sparse and incomplete due to the time of writing). Likewise, the issue category of ‘morality’ is lost by the Republicans by 2008–2012, and only becomes a Democratic strength (ranked second) when it is a bottom-ranked issue for the Republicans by 2012–2013. In the United Kingdom, labour and employment was the Labour Party’s top-ranked issue in 1974–1979, and its third best in 1979–1983, but by 1983–1987 had become the Conservative Party’s third-ranked issue and then its second best in 1987–1992 and 1992–1997. By 1997–2001, however, the Conservatives had lost ownership of the issue and by 2001–2005 it was as low as tenth. Apart from these examples, there are no instances where one party loses an issue and another party gains it: there are no sudden and permanent realignments. It might be the case that there is a considerable lag between the loss of an issue and the gain of an issue by another party (i.e. one spanning beyond our final data points), but the absence of a clear realignment in issue ownership is indicative of the nature and basis of ownership. It is not possible for a party to own an issue just because another party does not. We suggest that ownership is earned rather than conferred by the relatively poor rating of another party. This is an important reason to use a within-party measure of ownership. It makes possible the observation that reputations on issues are not symmetric and enables better understanding, therefore, of the causes of these relative shifts.

There is an interesting additional observation we can draw from Tables 3.2 and 3.3, where we see that two parties can be ranked relatively highly on the same issues, but one party can lose ownership. It is difficult to draw too much from the earliest period when the number of issues is smallest, but it appears that both the Republicans and Democrats were ranked best on the issues of the economy, defence and foreign affairs. These issues are later lost by the Democrats, after the mid-1970s, and retained as the Republican’s best issues, with the exception of a dramatic loss in foreign affairs rating for the Republicans during the 2012 to 2013 period. This highlights that voters do not necessarily always associate any party with an issue, i.e. there may be certain issues where many voters are unsure who is best, and thus both parties in a system receive low ratings.

The next observation from Tables 3.2 to 3.9 is that when an owned issue is ‘lost’ it appears to be lost irrevocably. Of course, the data do not last indefinitely, and issues could become owned by a party in the future. But Tables 3.2 to 3.9 only reveal a few instances where an issue moves out and then in from the list of top-ranked issues. One such instance is the category of ‘rights and minorities’ for the Democrats, which covers civil rights and race relations. This appears to be an issue that tends to be owned by the Democrats, but it is lost for two periods (becoming a middle-ranked issue); in the 1950s to 1960s (during the 1956–1960 and 1964–1968 election cycles) and then again from 1988 onwards (only then becoming a second-ranked issue during the 2008–2012 cycle). This might be an issue set that the Democrats have a propensity to be rated best on, but this does not insulate them from a loss of reputation. Indeed, inspection of the underlying poll data (for the earlier period not reported in Figure 3.2) indicates that while the Democrats still held a considerable lead on the issue of rights and minorities over the Republicans during the 1950s and 1960s, the party’s ratings declined relative to other issues. The issue of immigration offers a similar pattern for the Conservative Party in the United Kingdom. This issue is an owned issue which switches quite significantly from being second-ranked (1966–1970) to fifth (1970–1974), to second again (1974–1979). Again, this reveals how an issue for which a party has a very strong reputation can still cease to be an owned issue in certain cases, even if it later returns to being an issue asset. Another example comes in the form of foreign affairs for the German Christian Democrats. This high-ranking issue for the party (placed second or third between 1957 and 1972) is lost to a middle-ranked issue between 1972 and 1987, but becomes the top-ranked issue for the CDU during the 1990–1994 election cycle after reunification.Footnote ⁷

The final point to note from Tables 3.2 to 3.9 is that the contestable notion of ownership is again apparent. While we reveal within-party rankings of issue competence ratings and classify a drop in rankings as a loss of ‘ownership’, a range of other nuances are evident. For example, we might think of ownership as a tendency for a party to be rated positively on an issue; in such cases the return of an issue into a party’s most highly ranked issues, some years later, might suggest its reputation is more durable over a long time period. Another might be the loss of a positively ranked issue that an opposing party never comes to own. This conceptualisation would not, however, allow for explanation of the many losses and gains in issue reputations that so rarely result in realignments or switches of issue ownership. Most importantly, we reveal important and substantial variation in issue reputations across issues, parties and countries. These require explanation, and they relate to the concept of issue ownership – as it has been understood to date. However, they challenge a notion that ownership is always relative to a rival party. In our data, an issue can be ‘lost’ even if there is the possibility the party may still have a higher rating on the issue than its rival.

Issue Ownership and Partisanship

If issue ownership relates to long-term reputations of parties on issues, then it may be less prone to a problem of endogeneity than other evaluations in public opinion. This follows from the explanation of ownership resulting from long-term commitments to issues and from the representation of constituencies. These are characteristics which may be recognised and credited to the party by a majority of voters, rather than only the natural supporters of the party. It also follows from an explanation of ownership loss arising from a major and symbolic policy failure, which causes voters to re-evaluate and assess a long-standing view of the party as more committed to and trusted on an issue. Such a reputational loss may be recognised by supporters of the party, cutting through favourable partisan perceptions.

This may help to define the basis of issue ownership. If a party has a lead on an issue over other issues, it is likely to be the better rated party by a majority of the electorate; attracting the support on the issue from its own supporters, from floating voters or independents, and perhaps even from the supporters of rival parties. If the Republican Party is evaluated on a Democratic-owned issue, it will mainly be Republicans who rate it as best on that issue. But if the Republican Party is evaluated on an owned issue, their reputation for better handling will most likely extend beyond their partisan base. This does not mean that a party will be assigned positive evaluations on any issue by a rival party supporter, but that a party’s reputations on issues will follow the same ordering among rival supporters.

We cannot break down our aggregate data in Figures 3.1 to 3.5 and Tables 3.2 to 3.9 by partisanship. This requires individual-level data with consistent and repeated questions about issue handing.

The rolling cross-sectional survey of the National Annenberg Election Study (NAES) allows us to compare how US partisans (and independents) evaluated the incumbent George W. Bush and prospective party nominees (Hillary Clinton, John Edwards, Barack Obama, Rudy Giuliani, John McCain, Mitt Romney, Fred Thompson) for the 2008 presidential election for their handling of issues in the pre-primary season, between September and December 2007. The question asked on presidential issue handling was: ‘Do you approve or disapprove of the way George W. Bush is handling this issue: [… the economy]?’ The question asked on candidates was, ‘Are there any candidates who you think would do a particularly good job of handling [the nation’s economy]?’ These presidential/candidate issue-handling evaluations are not direct measures of party reputation, but it is reasonable to expect that candidates are evaluated at least partially on the basis of the issue-handling reputations of their parties.

The following graph compares the average issue-handling scores for President George W. Bush on seven issues, covering a period of 16 weeks in total. We expect (a) Republicans to rate George W. Bush with higher mean ratings than Democrats, but (b) the ranking of issues to be similar. We can analyse this by presenting the mean issue evaluations by Republican Party identifiers (respondents who indicated that they are either strong, weak or lean Republican) and the mean issue evaluations by Democratic Party identifiers (strong, weak or lean Democrat) over time. This is presented in Figure 3.3. We also compare the ranking of George W. Bush on each issue for Republican Party identifiers, Democratic Party identifiers and non-identifiers,Footnote ⁸ taking the mean rating over the time period. These rankings are presented in Table 3.10.

Figure 3.3 Republican and Democrat identifier ratings of President George W. Bush on issues

The differences in Figure 3.3 are striking. They reveal the way in which mean presidential handling evaluations are strongly associated with partisanship. If we use the average evaluations of candidates (or parties), or the lead of one party over another, we can expect strong effects of party identification, either leading to party identification or arising – endogenously – because of party identification.

However, Table 3.10 reveals that the relative issue reputations of George W. Bush are largely the same, revealing very little or no systematic partisan bias or influence. This pattern is consistent whichever candidate for the party presidential nomination we analyse using the NAES rolling cross-sectional data for the period between October and December 2007. In fact, where we have fewer issues available (four), we find exactly the same rankings (in evaluations of issue handling for Hillary Clinton, John Edwards, Barack Obama, Rudy Giuliani, John McCain) and almost identical rankings (Mitt Romney, Fred Thompson). (See the Appendix, Tables A3.3 to 3.9.)

We can also check to see whether these consistent rankings are unique to the United States and to evaluations of presidential (candidate) handling, and further whether they remain consistent assessed over a longer time period and using survey questions with different question stems and response options.

In Britain, the CMS provides the most useful data, asking respondents to evaluate the handling of issues by governing parties, rather than candidates. Specifically, for the incumbent Labour Government it asked ‘How well do you think the present government has handled the National Health Service?’, and for the opposition Conservative Party ‘How well do you think a Conservative government would handle the crime situation in Britain?’ The data can be disaggregated by party identification for the same issues (with handling of the financial crisis added to the set of issues in October 2008) over time.

The CMS data run over a substantially longer time period than the NAES. This offers a total of 71 monthly cross-sectional observations over the period between April 2004 and April 2010.Footnote ⁹ We examine average party issue-handling evaluations (where the scale runs from 0 being equal to ‘very badly’ to 4 equal to ‘very well’) across respondents who identify (either very strongly, strongly or not very strongly) with the Conservatives or with Labour, and for people with no party identification. With these data it should be harder to find consistent relative issue-handling rankings across partisans and independents.

Figure 3.4 reveals the mean scores for each issue for Conservative identifiers and Labour identifiers, and Table 3.11 presents the rankings.

Figure 3.4 Conservative and Labour identifier ratings of the UK Conservative Party on issues

Figure 3.4 reveals partisan differences in mean levels that are substantively identical to those shown for President George W. Bush in Figure 3.3. The mean evaluations are strongly associated with the partisanship of respondents, as we would expect. However, as with the US data, Table 3.11 reveals substantial similarity between Conservative and Labour party identifiers in the average rankings of party issue evaluations over the period. There is slightly weaker correspondence across all issues than evident in the NAES for presidents and candidates, which could arise from the longer time period (which allows for more variation), the evaluation of parties rather than candidates, the different question wordings or from country differences. However, Table 3.11 reveals – with the exception of one issue, terrorism – that the same issues feature in the top, middle and bottom rankings for Conservative partisans and for Labour partisans.

From these analyses, we suggest that the ranking of parties on issues provides a measure that is far less susceptible to partisan bias or partisan effects than the mean level or the between-party lead.

Figures 3.1 and 3.2 earlier showed how parties tend to have issues on which they are consistently stronger or weaker, how public opinion fluctuates on each issue and how public opinion also moves in a common general direction across issues, indicative of generalised assessments of competence. We now consider evidence for issue-specific performance fluctuations.

The Characteristics of Performance Fluctuations

Here we assess whether changes in issue ratings are distinct to issue ownership by considering whether there is greater, lesser or an equal amount of variation or volatility in public evaluations of parties on issues, irrespective of whether an issue is owned or not owned. This is examined by comparing the variance of public assessments of performance on individual issues to their rank. If issue ownership is conceptually distinct to the fluctuations in public opinion about competence we should expect a similar amount of volatility in public opinion on a party’s ‘best’ issues relative to worse. This would tell us that ownership, as a relative ranking, is the more stable concept, but that ongoing fluctuations and change in assessments should be conceptualised independently.

We display box plots in Figure 3.5 of mean ratings of the Republican Party on all issues over the entire period 1939 to 2013. The box area indicates the interquartile range, while the line that intersects it denotes the median value of mean issue-handling ratings. The ‘whiskers’ that protrude from the top and bottom of the box indicate the minimum and maximum values (with any outliers marked with a dot).Footnote ¹⁰ We are interested in whether variation in public assessments of performance on individual issues reflects the relative strength (or weakness) of parties on an issue. We would find evidence of lesser variance on owned issues if the range were lower around issues exhibiting higher mean values. As shown in Figure 3.5, this does not appear to be the case.

Figure 3.5 Box plot of issue ratings of the US Republican Party, by year

In the figure, we see variation in the median level of issue competence for the Republicans across issues, for example being viewed more positively on defence than on the environment. There are also differences in the size of the interquartile range across issues, with a much wider spread of values for defence than immigration. However, the figures reveal no systematic relationship between higher rated issues and wider or narrower ranges of values. Indeed, we observe high means with large ranges more often than we observe high means with small ranges, or low means with high ranges. That is to say, there appears to be no systematic variation between mean and range which would be consistent with higher rated issues exhibiting less short-term variability.

To ensure these findings are not party-specific we examined the same relationships focusing on the US Democrats, the UK Conservative Party and the UK Labour Party. In none of those box plots did we find a clear relationship between the overall mean level of public opinion and the degree of variance around those estimates.

We can examine the relationships further by focusing on the mean and variance for different time periods. It might be the case, for example, that the greater range of values of Republican handling of defence, above, is partly a consequence of being observed at different points over a much longer time period (1940 to 2013) compared to handling of an issue such as the environment (1979 to 2011). To test this idea, we compare the standard deviation of mean issue-handling ratings for a party on a given issue during each election cycle (a four-year presidential term in the United States and a parliamentary cycle in the United Kingdom) against the ranking of that issue in that election cycle for that party, and we display those figures for our four party cases (the Republicans and Democrats in the United States, and the Conservatives and Labour in the United Kingdom). Figure 3.6 displays the scatter plots of the standard deviation of issue-handling ratings against the ranking of that issue for the party in the same election cycle.

Figure 3.6 Scatter plot of standard deviation of issue ratings against rank (within election cycle)

We find virtually no difference between high- and low-ranked issues in the degree of variation observed (within election cycle), with all the R-squareds being close to zero and with little structure to the data. The slight upward slope for the Republicans, Conservatives and Labour might simply be a function of the fact that sampling error of surveys is a function of the sample proportion, so higher mean ratings would tend to be associated with slightly more variation simply as a product of statistical theory. However, even if these upwards slopes are due to greater variation in public opinion, we note that this is in the opposite direction than would be expected if there were greater stability on issues on which a party has its most positive reputation, or issue ownership. This confirms that short-term variability of performance assessments is not tied to long-term ownership of issues, or if there is a relationship, it is extremely weak (and in the direction of more variability on best rated issues). There is meaningful variation in performance assessments that is independent of the concept of long-term ownership.

Performance Fluctuations and Partisanship

Earlier we revealed how the relative rankings of parties on issues were held consistently by partisans as well as by non-partisans and rivals. Here we examine whether performance evaluations are likely to provide explanatory power above and beyond the tendency of partisans to evaluate a party more positively, i.e. whether partisans update evaluations of preferred parties and rivals on performance.

One view of partisanship may be that partisans are too biased to notice positive and negative performance. They neither receive nor accept a negative message about their preferred party (see Zaller’s acceptance and rejection theory Reference Zaller1992), and their positive (and negative) evaluations may already be established. Another view is that partisanship is itself a function of performance assessments over time, such that partisanship arises from positive performance and can be lost due to negative performance (Fiorina Reference Fiorina1981). This view of partisanship would expect that shocks carry forward, shaping party identification. It may also be the case that current partisans are particularly like to notice positive and negative performance and to respond to it. They are, after all, the same people who tend to give greater attention to politics. They may be particularly responsive to negative performance of rival parties and positive performance of their preferred party, alongside the bias in attribution of responsibility known to co-exist with partisanship (Tilley and Hobolt Reference Tilley and Hobolt2011). Independents, on the other hand, may exhibit opinion about politics that is noisier, with shocks to performance being less likely to structure future evaluations.

In the analysis that follows we examine the degree to which issue-handling evaluations carry forward the effects of past shocks – using the aggregate level NAES and CMS data discussed above – and how these patterns differ across partisans, rival partisans and independents.

Tables 3.12 to 3.14 report autoregressive distributed lag (ADL) models of issue-handling evaluations of President Bush, presidential nomination candidates and parties (pooling data across issues by president/candidate/party). The ADL takes the form Y_it = α₀ + α₁Y_it−1 + β_i + ε_t, where Y_it refers to the competence evaluation at time t for a given issue i, and Y_it−1 refers to the competence evaluation on the same issue in the previous time period. We include issue-fixed effects (β_i) to control for variation in the average handling rating by issue. In this model specification, the parameter α₁ indicates the rate of autoregression. That is, the degree to which current competence evaluations are a function of competence evaluations in the previous period.Footnote ¹¹ Of key interest is the strength of the autoregressive process which takes a value between 0, where no portion of past evaluations are carried forward, and 1, where all shocks to performance evaluations are remembered. Analysing this autoregressive process for each group reveals the extent to which the persistence of performance evaluations varies across partisans, rival partisans and non-partisans.

In Table 3.12, we consider performance evaluations of President George W. Bush across seven issues on a week-by-week basis over the 16-week period using data from the NAES. These are the best data we have disaggregated by partisan identity for the United States.

These results reveal that non-partisans and rival partisans exhibit much more short-term fluctuation in their issue competence assessments than partisans, who were much more inclined to update their evaluations of George W. Bush in response to shocks. The coefficient of the autoregressive term (Evaluations_t−1) is equal to 0.2 (p<0.001) for partisans but not significant for rival partisans or non-partisans. The goodness-of-fit of the model, as measured with the overall R-squared, is 0.7 for partisans and 0.9 for rival partisans (perhaps revealing the highly persistent attitudes that Democrats held towards George W. Bush by this period), but equal to 0.1 for non-partisans.

We see a similar pattern for performance evaluations of prospective party presidential nominees in Table 3.13, but here we find a similar autoregressive process for partisans as for rival partisans. The pattern is of significantly greater persistence among partisans and rival partisans compared to non-partisans. Note that the R-squared for non-partisans is much lower than that for the models for partisans and rivals. This indicates that past competence shocks provide far less information about current evaluations for non-partisans.

The differences we observe between Tables 3.12 and 3.13 – that of substantial updating for rival partisans for candidates, but only for the partisans of the president’s party for presidential handling – may arise due to the highly partisan and competitive period in which presidential candidate data are asked: the candidate campaigns for each party’s presidential nominee.

The analysis is again repeated using the UK data, in Table 3.14. This uses performance evaluations of the Labour government and Conservative opposition across seven issues over the period between April 2004 and April 2010.Footnote ¹² These data provide monthly units of analysis rather than weekly, which might be expected to exhibit higher rates of persistence.

In Table 3.14, we find stronger persistence for non-partisans in the United Kingdom than we found for the United States and stronger rates of persistence overall: for partisans and rivals. However, once again we find that partisans are much more likely to exhibit updating, in terms of persistence, than are independents. On a month-by-month basis, partisan and rival partisan performance evaluations are substantially autoregressive, with the coefficient of around 0.65 in both cases (p<0.001), whereas non-partisan evaluations are more weakly dependent, with a coefficient of closer to 0.3 (p<0.001). The R-squared is again much lower for non-partisans compared to partisans and rival partisans.

These UK and US differences may arise due to the stronger party system in Britain, which results in stronger party-based assessments across the board, or the differences between evaluations of individuals (presidents and candidates) and parties, but more likely and most simply, probably arises as a consequence of using the monthly unit of aggregation instead of weekly observations. However, we again find much weaker persistence among non-partisans than among partisans, consistent with the US findings. And as in Table 3.13, our results demonstrate persistence in evaluations among partisans and rival partisans alike.

Our analyses confirm that evaluations of performance by partisans or rival partisans are subject to long-term accumulation of short-term performance shocks, whereas evaluations of non-identifiers are only temporarily affected by recent events or performance, exhibiting little ‘memory’ of past shocks and fluctuating noisily. Partisans notice and respond to past information, exhibiting more performance updating than non-partisans. To the degree, therefore, that performance fluctuations matter, they should matter just as much, if not more, for people who identify with a party than for people with no partisan affiliation.

The Characteristics of Generalised Competence

As highlighted earlier in Figures 3.1 and 3.2, there are periods where we can discern a shift in public opinion about the competence of parties where those translate across policy issues. Crucial to understanding these periods is whether these are simply a function of increasing or decreasing popular support for a party, which has the effect of increasing or decreasing issue-handling evaluations (whether generalised shifts in competence are endogenous to popularity). Another reason might be the transfer of competence ratings from one issue to another, such that a party that is seen to fail significantly on an issue, or conversely to demonstrate significant success, might be judged more negatively – or positively – for its competence on other issues. This is the notion behind our work on ‘macro-competence’ (Green and Jennings Reference Green and Jennings2012a), which reveals how common variance exists in public opinion about competence.

The idea behind macro-competence is that voters use heuristics to assess the competence of political parties. This allows low-information voters to make boundedly rational judgements about party competence; drawing on small amounts of information to update their assessments of performance by transferring evaluations across issues or using a generalised sense of the performance of parties.

We reveal here how much shared variance exists in public opinion about competence in countries for which we have sufficient aggregate level survey data over time: the United States, United Kingdom, Canada, Germany and Australia. Stimson’s (Reference Stimson1991; Stimson et al. Reference Stimson, MacKuen and Erikson1995) ‘dyad ratios algorithm’ is used to extract the underlying dimensions of citizens’ evaluations of party competence across all available survey items. This method builds on the idea that ratios of aggregate-level survey responses to the same question, asked at different points in time, provide meaningful information about the relative state of public opinion (see Stimson Reference Stimson1991 and Bartle et al. Reference Bartle, Dellepaine-Avellaneda and Stimson2011 for an extended discussion of the method). Aggregate competence evaluations can be scored either as the raw percentage of respondents naming a party as most competent/trusted on an issue in a survey (e.g. 55 per cent rating the Republicans as best on defence), or as an index of the relative proportion of respondents naming either of the main parties as the most competent/trusted (e.g. the ratio between 55 per cent rating the Republicans as best on defence and 35 per cent rating the Democrats as best on the issue, so calculated 0.55 / (0.55 + 0.35) = 0.61). We use the former measure here since it allows party competence evaluations to vary due to changing levels of respondent uncertainty and favouring of other parties (or ‘none of the above’).

Each survey item can be expressed as the ratio of competence evaluations at two points in time, i.e. a ‘dyad’. The ratio provides an estimate of the relative perceived competence of a party, on a given issue, in years t+i and t+j:

$C_{ij}=\frac{X_{t+i}}{X_{t+j}}$

This enables recursive estimation of the competence index for each survey item for each time period (i.e. years or quarters) based on all data available for that time period (and other time periods). There are multiple overlapping estimates of these separate competence indices and each one is not an equivalent indicator of the underlying construct. To solve this, the dyad ratios algorithm iteratively estimates the squared correlation of each series with the latent dimension and uses this to weight the series, proportional to their indicator validity (Bartle et al. Reference Bartle, Dellepaine-Avellaneda and Stimson2011: 269).Footnote ¹³ The method extracts the central tendency of all survey items relating to evaluations of party issue competence, analogous to a principal components approach. Note that these measures do not parse out the influence of electoral popularity or partisanship (see Green and Jennings Reference Green and Jennings2012a: 335), and we turn to those relationships later. They simply reveal how much common variance exists across the time period and the parties for which we have data. This is derived from 5,436 survey items for the United States, 3,536 for the United Kingdom,Footnote ¹⁴ 1,120 for Germany, 752 for Australia and 160 for Canada (consisting of 11,004 items in total).

Table 3.15 reports the proportion of variance loading onto the first and second dimensions of the extracted measure of macro-competence, and the mean and standard deviation of each dimension.

Table 3.15 reveals particularly high proportions of variance loading onto the first dimension, or ‘macro-competence’. This first dimension represents the greatest proportion of common variance in competence evaluations across all issues. It is interpretable as the central tendency in the public’s evaluations of the issue-handling capabilities of parties. Higher values of macro-competence indicate that the public views a party more positively in its handling across a range of issues. Lower values indicate that the public tends to view a party as handling issues poorly, in common. The remainder of variance explained – approximately 35 per cent in the United States, 30 per cent in the United Kingdom, 15 per cent in Germany and 25 per cent in Australia – provides an assessment of the unique variation that is specific to individual issues; i.e. the part that can be explained by issue-specific changes in public opinion about performance.Footnote ¹⁵ The first and second dimensions together show very high proportions of shared variance.

The amount of shared variance is an important observation in its own right; it suggests that generalised competence assessments provide an additional and substantively important concept in the understanding of public opinion about competence. A generalised concept of competence is distinct to the concept of issue ownership, and it is distinct to measures of issue-specific fluctuations in policy performance. This common variation may also indicate a strong partisan component, such that the common variation is driven by shifts in party attachments and popularity. This is also an important insight, insofar as a small but still very significant proportion of variation in public opinion about competence is not shared across issues.

Generalised Competence and Partisanship

Here we address the question: to what extent are generalised competence evaluations a function of partisanship, and to what extent do generalised competence evaluations shape changes in partisanship? We assess these questions in two ways. First, we visually present the relationship between macro-competence and partisanship over time. We have aggregate-level data on partisanship (‘macro-partisanship’) in the United States, United Kingdom and Germany, and so time series for these variables are displayed for those countries. Second, we assess the temporal relationships between macro-competence and partisanship using tests of ‘Granger causation’ (Granger Reference Granger1969).

Figures 3.7 to 3.12 display the macro-competence and macro-partisanship series for each party by country. The macro-partisanship series represent the percentage of respondents indicating political affiliation to a particular party (e.g. the Democratic Party in the United States or the Social Democrats in Germany). In the United States and United Kingdom, we have multiple sources of survey data concerning party identification, so we again use Stimson’s dyad ratios algorithm to create a single index (following Erikson et al. Reference Erikson, MacKuen and Stimson2002). For Germany, we have a single data source, Forschungsgruppe Wahlen’s monthly Politbarometer, so we simply take the average percentage using these data.Footnote ¹⁶

Figure 3.7 Macro-competence and macro-partisanship, US Democratic Party

Figure 3.8 Macro-competence and macro-partisanship, US Republican Party

Figure 3.9 Macro-competence and macro-partisanship, UK Conservative Party

Figure 3.10 Macro-competence and macro-partisanship, UK Labour Party

Figure 3.11 Macro-competence and macro-partisanship, German Social Democratic Party

Figure 3.12 Macro-competence and macro-partisanship, German Christian Democratic Union

It is not possible to assess the precise temporal relationships between macro-competence and macro-partisanship by visual inspection of the figures alone, but it is evident that while there are clear parallels between the measures there is also a substantial degree of difference. In the United States, for example, we see that macro-competence and macro-partisanship have tended to move together over time, but have diverged for both Republicans and Democrats since around 2000. In the United Kingdom, the series have also tended to track one another, but have diverged at points in time. While the macro-competence of the Labour Party declined after 1997, its macro-partisanship remained relatively stable. In Germany, macro-competence tends to fluctuate rather more than partisanship, which has declined since the 1970s. Thus, while it is evident that generalised evaluations of party competence and partisan attachments are related, they are not synonymous in their levels or variances.

Next we examine the temporal interrelationship between the series. This helps us establish whether changes in macro-competence tend to precede changes in macro-partisanship (performance-based updating) or whether changes in macro-partisanship tend to lead changes in macro-competence (partisan conditioning). Specifically, tests for Granger causation (Granger Reference Granger1969) determine whether past values of a variable x improve prediction of another variable y, relative to prediction from y from past values of itself alone. This is not a test of causation in the strictest sense, but evidence of the predictive content and the temporal ordering of one measure in relation to another (Granger Reference Granger1988). If macro-partisanship shapes macro-competence (and if macro-competence were endogenous to macro-partisanship), we would at least expect the temporal ordering between the variables to tend to be stronger in the direction from macro-partisanship to macro-competence than from macro-competence to macro-partisanship.

Tables 3.16 to 3.18 reveal the tests for Granger causation between macro-competence and macro-partisanship (and vice versa) for each party, country and period where we have quarterly data series, which enable a more fine-grained examination of the dynamics of each of themeasures than we can obtain using annual data. Using shorter (quarterly) time intervals provides more information for analysis of the predictive content of lagged values of macro-competence and macro-partisanship: not only does this provide a larger N but it also reduces the likelihood that meaningful variation will be aggregated away (see Granger Reference Granger1969: 427). It does, however, mean that we are limited in analysis of Granger causation for periods where the data underlying the macro-competence and macro-competence series are quite thick.Footnote ¹⁷

These Granger causation tests reveal that macro-competence tends to lead macro-partisanship to a greater extent than macro-partisanship leads macro-competence. In just one instance – for the Republican Party in the United States – there is evidence of Granger causation in both directions. The estimated χ² test statistic considers whether macro-competence Granger-causes macro-partisanship or whether macro-partisanship Granger-causes macro-competence, such that where the χ² test statistic is significant there is Granger-causation between the variables. For example, in the case of the US Democratic Party, the χ² test statistic provides support for the hypothesis that macro-competence Granger-causes macro-partisanship (13.2), significant at the 99.9 per cent confidence level. In contrast, the χ² test statistic in the opposite direction (1.9) is not statistically significant at the 95 per cent level.

Our findings, on Granger causality, suggest that partisanship is not strongly exogenous to competence evaluations, but the temporal ordering is consistent with ‘running tally’ expectations of party attachments. This does not mean that party identification is not also acting as a perceptual screen. What we find here, and in previous analysis (Green and Jennings Reference Green and Jennings2012a), is that generalised competence evaluations are informative about subsequent changes in party identification, arguably as voters incorporate perceptions of competence into their affective orientations towards parties. In our previous work (Green and Jennings Reference Green and Jennings2012a), we found evidence that there is a strong and substantively large effect of macro-competence on vote intention, controlling for macro-partisanship and other variables. Our evidence suggests that macro-competence is far from purely endogenous to partisanship. Our later analysis (in Chapter 7) also bears this out, with more comprehensive models of competence. The concept of generalised competence evaluations provides a new and important addition to our understanding of public opinion about competence, and one which is conceptually and empirically distinct to partisanship.

Conclusion

There are three distinct empirical concepts to public opinion about competence: (1) The relative strengths and weaknesses of parties on different issues: issue ownership. These ‘level differences’ exhibit stability but also important variation, as exhibited in this chapter using decades of public opinion data from the United States, United Kingdom, Germany and Australia. The loss and gain of issue ownership is a topic that requires much more investigation, but it is only made possible by careful and accurate conceptualisation of the key variance that needs to be explained. The next chapter offers a theory and evidence to account for these losses and gains. (2) Fluctuations in public opinion about party competence on specific issues. These short-term changes in issue-based ratings are independent of within-party ownership rankings; owned and non-owned issues exhibit similar amounts of variance over time. (3) Generalised competence evaluations, which account for the largest proportion of variation in public opinion about competence and which represent the large shifts in direction in overall competence across the issue agenda. These three concepts are empirically and conceptually distinct.

This chapter also dealt with the relationship of each of the three concepts with partisanship. In each case we revealed how the concept is distinct to party identification.

Following our measurement of issue ownership as the ranking of better to worse within-party evaluations, we are able to overcome some potential problems arising from endogeneity. Partisans are more likely to rate their party higher on all issues, but the within-party ranking is broadly equivalent, irrespective of partisanship. Reputations are reputations, and they cut through the potential for a partisan bias or lens. We are concerned about measures of issue ownership that rely on the average of issue-handling evaluations for parties, or the gap between parties. As revealed here, there is a very strong relationship between party identification and mean party or candidate ratings on issues, which we assume reflect partisan conditioning to a large degree. However, a measure of issue ownership which uses party reputations as its basis, and which analyses the relative issue-handling strengths and weaknesses of a party, provides consistent evidence which is far less sensitive to partisanship or not sensitive to partisanship at all.

With respect to performance fluctuations, we identified strong effects of past ratings on partisans’ ratings of political parties for competence, but a much weaker autoregressive tendency in the views of non-partisans. That is to say, we believe performance fluctuations reflect meaningful updating among those people who have a prior partisan attachment.

There is a close correspondence between macro-competence and macro-partisanship but these concepts are by no means synonymous. Testing for the temporal ordering of both concepts across countries reveals that generalised competence assessments more commonly lead partisanship than vice versa. The implications may be important for our understanding of partisanship as a running tally of performance and they suggest that macro-competence is not simply the sum of over-time movements in partisanship.

The literature has been inconsistent in the use of terms and concepts in the study of issue ownership, competence and valence. This chapter demonstrates three ways forward in understanding public opinion about competence and in measuring meaningful and consequential concepts: issue ownership, issue-specific performance fluctuation and generalised assessments of competence.

What are the implications of three distinct concepts in public opinion about party competence? The answers to this question are sought in the remaining chapters of this volume. Given the frequency and potential significance of changes in within-party issue ownership, we set out a theory to explain these changes. This is the subject of Chapter 4. In Chapter 5 we examine the way fluctuations of performance assessments correspond to the performance of governments, and therefore relate to governing party vote choices much more than oppositions’. Chapter 6 analyses the concept of macro-competence in much greater detail, revealing the systematic trends in these generalised competence evaluations and how those cycles offer an explanation for the ‘costs of governing’ experienced so regularly by incumbents. Chapter 7 draws the three concepts together and reveals how each exerts an independent effect upon party choice, as well as considering when competence is more – and when it is less – strongly associated with party support.